Choose the right source for each claim

Your assistant knows your tastes, but not tonight’s menu

You have a clear request from Module 1. Now imagine the assistant suggests a dish that was removed last week. The missing capability is grounding: connecting a claim to information that supports it.

Retrieval-augmented generation (RAG) means retrieving relevant material and supplying it to a model before it answers. In our project, a menu description can support what a dish contains. A live quote is needed for the current total. A cached description cannot establish either availability or the final price.

Check your understanding

A cached menu lists a dish at $16. What can you safely conclude?

Give evidence an address and a date

Attach a source identifier, an item identifier, and a checked time to evidence. “Menu A, item tofu-bowl, checked at 18:05” is more inspectable than “I read it somewhere.” A timestamp helps you reason about freshness; it is not a guarantee.

Separate relatively stable descriptions from fast-changing stock and charges. Our lab uses fictional, fixed records and marks stale ones. In a real product, define how long each source may be reused and recheck before a consequential action.

Check your understanding

Match each claim to its best evidence.

Make uncertainty visible in the recommendation

A useful recommendation can say: “This menu describes a tofu bowl at $16. I have not verified the current checkout total.” That is a narrower and more useful claim than “This will definitely arrive for $20.”

The Retrieval Lab implements filtering and ranking over sample data. It does not run an embedding model or a generative model. You will inspect which record supports each result before adding those components. Background: retrieving context with its meaning intact.

Check your understanding

Write a recommendation when the menu is known but the final charge is not.

Build checkpoint

Open the Retrieval Lab and follow the practice above. Each lab explains its inputs, its actual logic, and its limits. Save your observations below; your check-ins and reflection remain in Product Academy.

For an AI-assisted code change, download the lab bundle from the lab page. Ask your coding assistant to explain the relevant function first, change one behavior, and add one test that could fail. Compare the result with your prediction.

Save your experiment and check-in

Record your prediction, what you tried, what happened, and the evidence for your conclusion. Name one thing you can now do independently and one uncertainty to revisit.

Key takeaways

• Match each claim to the source that can support it.
• Retain source IDs and freshness information.
• Separate menu descriptions from current operational facts.